Research On Waveform Classification Method Based On 3D Seismic Data

With the progress of the society, the natural energy exploration is becoming more and more important in our daily life. And with conventional oil and gas reservoirs are becoming exhausted, the demand for unconventional reservoirs energy exploration is rising, and promotes the development of seismic exploration technology. As an important method to analysis the underground strata tectonic and reservoir distribution, waveform classification technology is also obtained fast development, and has become an important part of the energy exploration.This paper first reviewed the research background and significance of the waveform classification technology, and summarized the problems in waveform classification technology of three-dimensional(3D) seismic data. Based on three aspects of 3D seismic data preprocessing, feature selection and classification method, we studied deeply that how to solve the problems existing in waveform classification technology applied in 3D seismic data, and developed a set of waveform classification methods which appropriate for the actual seismic data processing. The specific works of the innovation are as follows:1. According to the seismic data characteristics of waveform classification applied in 3D seismic data, we proposed a new waveform classification method based on morphology applied in 3D seismic data in this thesis. First, the waveform integration method based on morphology which helps to eliminate the influence of noise and to retain the waveform similarity between 3D seismic signals, was introduced into the 3D seismic signal data preprocessing for subsequent processing to provide reliable basis. Second, aimed at eliminating the influence of the errors of seismic horizon interpretation in waveform classification, singularity detection method based on wavelet transform was introduced into the processing of horizon error elimination. Finally, classification and identification of seismic signal was processed. Due to waveform features retention and horizon errors elimination, the method improved the accuracy of waveform classification results, and provided reliable basis for the subsequent seismic facies analysis.2. In traditional waveform classification methods, unsupervised classification methods were always used, which neglected the important information of well logging data. This influences the classification accuracy, and the contact of the actual category meaning and classification result is not close enough. For the problem, we presented a supervised waveform classification method based on support vector machine applied in 3D seismic data and the specific process. The method introduced an improved genetic algorithm in the process of feature selection to optimize the characteristics, reduced the redundancy of characteristics data and improved the classifier performance. At the same time, considering the well logging data distribution with small sample and nonlinear, the support vector machine(SVM) algorithm was introduced into the classifier design of 3D seismic signal. This method used the important information ignored by traditional unsupervised classification method, improved the resolution and accuracy of waveform classification results, and the classification results had been clear about the actual meaning category.The proposed waveform classification methods were applied to waveform analysis of many realistic 3D seismic data survey. From the waveform classification effect of actual survey, identifying geological details such as river, fault characteristics and waveform classification efficiency, etc., are better than the traditional waveform classification methods applied in seismic data.